Nobody knows what a corporate video returns. So we measured 3,053 of them.

Why this number does not exist anywhere

Before the numbers, the obvious question: an industry has been selling corporate films for two decades, so why is there no benchmark for what they achieve? The answer is not conspiracy. It is that nobody in the chain is paid to produce one.

The production companies are paid for the film, not for its performance. Their case studies show the work, never the view count six months later — and a studio that published median performance across its whole client list would be the only one doing so, at its own expense. The clients mostly do not measure, because the film was commissioned for an occasion rather than a target, and an anniversary has no KPI. The platforms hold the data but have no reason to slice it by a genre that only exists in a marketing brief.

So the number sits in plain sight, on hundreds of thousands of public video pages, and nobody adds it up. That is the entire novelty of what follows. There is no privileged access here, no leaked dataset, no survey panel — just the public view counter under every one of these films, collected on one day and counted honestly.

What exactly was measured

There is no official statistic on corporate films anywhere. What exists is agency pages with claims on them. So I counted, using a rule that is deliberately blunt so nobody — including me — can bend it after the fact.

Included: every public YouTube film that describes itself, in its own title or description, as an Imagefilm, Imagevideo, Unternehmensfilm, Unternehmensvideo, Firmenfilm, Firmenporträt or Unternehmenspräsentation. Those are the German trade terms for a corporate film. I do not decide what counts as one — the person who uploaded it does. The search ran across 115 queries: the bare term, the term combined with 50 industries, and with 20 regions and cities.

Removed: how-to and pricing explainers, showreels and portfolios belonging to production companies, public broadcasters and government bodies, and anything under 30 seconds or over 30 minutes, because those are teasers and recordings rather than films. What remained: 3,053 films on 2,230 channels.

What these numbers do not say

Three limits, which I would rather state myself than have pointed out. One: only YouTube was measured. If your film also plays on your home page, on LinkedIn, at a trade-fair stand or in a reception area, that reach does not appear here. Two: a film never uploaded publicly is invisible to this survey — the quietest cases are missing, which makes these numbers flattering rather than harsh. Three: views are not a business result. A film with 300 views, two of which were the right buyer, may well have paid for itself.

What the numbers do say is the order of magnitude at which public attention for this format actually happens. That order of magnitude is two to three zeroes smaller than the proposals written to produce it.

Individual companies are not named in this piece where their numbers are weak. They did nothing that warrants being named — they bought what they were sold. Every breakdown here is an aggregate.

844

The mean across all 3,053 films is 8,303 views. That is the number you would put in a deck, and it is misleading. The median — half above, half below — is 844. The mean sits almost ten times higher than the middle because a handful of very large films drag it there.

Half never reach a thousand

In plain shares: 13.3 percent of corporate films have fewer than a hundred views. 38.7 percent stay under five hundred. 54.0 percent — more than half — never reach a thousand. And 1.7 percent pass a hundred thousand.

That 1.7 percent is why the industry can talk about reach in good conscience. It is also why the arithmetic fails for everyone else: the top ten percent of films hold 85.9 percent of all views in the corpus. The top one percent holds 48 percent.

A market in which one percent of the output carries half the return is not a market with quality differences. It is a market with a distribution mechanism that most buyers never bought. The film arrives. The distribution was never ordered.

What one viewer costs

For the cost side I am not using someone else's estimate. I am using the corridor I published and justified before this survey existed: a conventionally shot corporate film in Germany costs between 10,000 and 50,000 euros, with the weight around 25,000. The line-by-line breakdown is in what a corporate video costs.

Divided by the 844 views of the median film:

And in the bottom quarter

That is the middle. A film landing in the bottom quarter — 231 views — costs €108.23 per viewer at a 25,000 euro budget. In the bottom tenth, at 74 views, it is €337.84. For each individual human being who clicked far enough to see the opening titles. Not who watched to the end — who clicked.

The comparison with bought reach is uncomfortable and belongs here anyway: a video view purchased through paid distribution costs a few cents depending on the audience. That does not mean a bought view is worth what an organic one is worth — it is not. It means there is a factor of several hundred between the two orders of magnitude, and that this factor appears in no proposal I have ever read.

The honest objection is that a corporate film is not meant to reach many people, only the right ones. True — which is exactly why the next number matters more than this one.

The channel with 91 subscribers

The channel holding the median corporate film has 91 subscribers. Half of all channels in this survey — 51.1 percent — have fewer than a hundred. 78.8 percent have fewer than a thousand.

That alone explains the arithmetic. A film on a channel with 91 subscribers has no audience on the day it is published. It has an address. Anyone who is meant to see it has to be actively sent there — through the website, the newsletter, the sales team, an ad, a trade-fair stand. If that does not happen, the film stays exactly where it was uploaded and is found only by people who were already searching for the company.

Seven likes

The second quiet number: the median film has seven likes, a like rate of 0.67 percent of views. For YouTube that is not catastrophic, but in absolute terms it means that under the film a company paid a five-figure sum for, seven people raised a thumb. There is no discussion, no forwarding, no second wave. The film is a monument: it stands there, it is well built, and it does not move.

That is the precise description of the problem. A corporate film is almost always commissioned as an object — a thing that exists at the end, gets signed off, and is then simply present. It is almost never commissioned as a process: something that has to travel to specific people and do something to them. Both briefs cost the same. Only the second one has a recipient.

The line that has not moved since 2010

Sort the films by year of publication and it looks like a collapse: films from 2019 have a median of 1,449 views, films from 2025 only 265. You could build a satisfying thesis on that — the corporate film has run its course, attention has moved elsewhere.

The thesis would be wrong, for a reason that is easy to miss: a film from 2019 had seven years to collect its views. A film from 2025 had one. Divide views by years online instead, and the collapse disappears completely:

A corporate film has been collecting between 116 and 196 views per year since 2010, with a median of 146. 2011: 196. 2019: 130. 2024: 121. In between, cameras, editing software, platforms, formats and eventually half the production stack were replaced. The number was not.

This is the real finding of the survey, and it is harder to live with than any decline story. There is no golden age to return to, and there has been no technical advance that solved the problem. What has stayed constant is not the quality of the films but the way they are commissioned. And that really has barely moved since 2010.

Three explanations the data does not support

A survey is only worth something if it is allowed to kill the assumptions of the person running it. Here are mine, in the order they fell.

"Shorter is better"

This is the industry's best-selling piece of advice, and in this dataset it is false. The 763 films under 90 seconds have a median of 532 views. The 1,029 films over three minutes have 1,248 — more than twice as many. The median film overall runs 2:19. Split into finer classes, as in the chart below, the same pattern runs from 387 views under a minute up to 1,339 at four to six minutes — the same data, cut two ways.

Read that carefully: it does not mean length produces views. The likelier direction is the reverse — a company that commissions a long film has invested more, has more to show and is more likely to actively push it to people. What it does mean for certain: if you think your film failed because it was thirty seconds too long, you are looking in the wrong place.

"The production company buried it on their own channel"

My second theory: films that live only on the production company's channel and never on the client's do worse. It sounded plausible. Twenty percent of the corpus sits that way.

The data says: no difference. Median on company channels 838, median on production-company channels 861. The gap is noise. Where the film lives is irrelevant — because both places have equally little audience. That is not an acquittal for the production companies. It is a sharper charge: it makes no measurable difference who uploads the film, because nobody distributes it.

"It used to work"

The third was the most stubborn and it died in section 05: adjusted for time online, a corporate film has been collecting 116 to 196 views a year since 2010, median 146. There was no good old day. There was only more patience with the same numbers.

By industry

Grouped by the industry each film was found under, the picture contradicts most expectations. The leaders are not the sectors with the largest marketing budgets. They are the sectors with the most concrete things to put in front of a lens.

At the top: logistics, median 1,427 views, followed by metalworking and machinery. These are the industries with something to show that an outsider never otherwise sees — a sorting facility at four in the morning, a mill cutting a part from solid stock, a high-bay warehouse in motion. Footage like that works because it is interesting independently of who paid for it.

At the bottom: consulting and law firms at 601, electrical and technical services at 375 — €66.67 per viewer. These are the industries where the film inevitably consists of people in meeting rooms, handshakes and drone shots over an industrial estate. Not because the filmmakers are worse, but because the work itself is invisible. A company that makes nothing visible has to build the film out of something other than its operations, and almost never does.

Why this happens

I have sat through enough of these meetings to recite the sequence. A company decides it "needs a new film". The trigger is usually an anniversary, a trade fair, a website relaunch or a change of generation. A production company is approached, a mid five-figure quote comes back, it is shot, there are three rounds of notes, the film is signed off and uploaded.

One question never appears anywhere in that sequence: who exactly is supposed to see this, and how does it reach them?

It never appears because in that room it belongs to nobody. The production company is responsible for the film, not for its distribution — it is paid for the work, and the work is finished on sign-off day. The company commissioned a film because it wanted a film, not because it wanted to reach a particular group of people. The result is a cleanly produced object with no recipient.

The second mistake follows from the first and costs more: everything rides on one attempt. Twenty-five thousand euros, one film, one upload. In no other marketing discipline would anyone accept that. Nobody runs a single ad and draws conclusions about the channel. Nobody sends one email to a list. Only in film does the single big swing still count as professional — and section 02 shows how that swing lands on average.

The reason for the exception is economic, not artistic: shot conventionally, every additional version costs another shoot day, another crew, another round of approvals. Variety simply was not affordable. That constraint disappeared roughly two years ago — what follows from it is in section 10.

Who counted this

Over the past two years I have produced around 1,500 videos together with a group of AI creators. Between them they have collected roughly 500 million views. I am making that figure checkable rather than decorative: on request I will show which channels those views were earned on, with examples. What I will not do is name clients who did not ask to be named.

Set against this survey the comparison is unfair, and I will give you the result anyway because it makes the order of magnitude visible: all 3,053 German corporate films together hold 25.3 million views. That is one twentieth of what 1,500 videos collected in two years.

It is unfair because these are two different species. One is material built so that somebody chooses to watch it — with everything that involves: a hook, a format, platform mechanics, many attempts. The other is a film built so a company can show it. The difference is not good versus bad. It is being watched versus being screened.

And now the contradiction

I sell AI production. And if you let me choose, I take the camera.

As a child I started working through the IMDb list of the 250 best films from the top down. I got to about two hundred. I kept a notebook and wrote down, after each one, what it had done to me — not whether it was good, what it did to me. I have watched The Intouchables more than ten times. Truffaut's The 400 Blows, Tornatore's The Best Offer, Fellini, Sorrentino, Mr. Nobody, It's a Wonderful Life from 1946. I mention it not as a credential of taste but because it explains why 844 views make me angry rather than amused.

The best work I have done professionally had nothing to do with AI. Fifteen days and roughly 3,500 kilometres across Ukraine in a camper van we converted ourselves, my father driving — that is becoming a documentary (Once Upon a Time in Ukraine). And last August my wife and I walked about 300 kilometres from Porto to Santiago de Compostela in ten days: walk every day, film every day, one proper camera and a GoPro on the chest running from the first step to the last (Pilgrims). There is no generated image that competes with the material from those ten days.

AI is a tool I use to reach a goal. Film is something else, and it has been with me since I was twelve. Holding both at once is not a contradiction I need to resolve. It is the reason I can judge a film without having to sell you one.

One more thing, because it belongs to how this work happens: I do not like working alone. There are a great many gifted people, and my approach is to bring them together rather than to be the centre of it. Most arrive through people I know. Together there is a different momentum — you find solutions you would not reach alone, and you learn on the way. Different goals need different talents; my job is to bring the right talent to the right undertaking and hold the direction together.

What a corporate video can be in 2026

If the mistake sits in the brief rather than the camera, a cheaper production helps not at all. A badly addressed film for 3,000 euros is the same mistake as a badly addressed film for 30,000, only with worse lighting.

What actually helps is four decisions, all of which are made before the first shoot day:

Where AI actually changes the arithmetic

The fourth decision is the technical one, and disclosure belongs here: this is what I sell. The decisive advantage of AI production is not that one film gets cheaper. It is that the twentieth film costs almost nothing.

Shot conventionally, every further version costs another shoot day. That is why "many attempts instead of one" has been advice nobody could afford. In AI production the marginal cost curve is flat: once the look, the characters and the world are fixed, the next variant is a question of hours rather than shoot days. For the first time that makes affordable what every other marketing channel has done as standard for years — try, measure, replace.

Then there is the quality question, which was still a fair objection two years ago. Across those 1,500 videos we have worked in very different styles, and where it stands today is this: AI adverts and AI corporate films can be taken to a level where the conversation about the technique stops and only the film is discussed. That is the point at which it gets interesting — and the reason the first enquiries for it are now large enquiries.

Two things AI cannot and should not replace, and I would rather say so upfront: real people from the business and real customer testimonials. A generated employee is not a workforce, a generated customer is not a reference — and in Germany the second is not merely dishonest but legally attackable. Wherever your team, your machine or your customers have to be on screen, you shoot. Everything else can be generated.

What I bring — and how you can check each part of it

You should not buy anything from someone whose claims you cannot verify. So here are the three things I bring, each with the way to check it.